MAROKO133 Breaking ai: Which Agent Causes Task Failures and When?Researchers from PSU and

📌 MAROKO133 Hot ai: Which Agent Causes Task Failures and When?Researchers from PSU

Share My Research is Synced’s column that welcomes scholars to share their own research breakthroughs with over 1.5M global AI enthusiasts. Beyond technological advances, Share My Research also calls for interesting stories behind the research and exciting research ideas. Contact us: [email protected]

Meet the authors
Institutions: Penn State University, Duke University, Google DeepMind, University of Washington, Meta, Nanyang Technological University, and Oregon State University. The co-first authors are Shaokun Zhang of Penn State University and Ming Yin of Duke University.

In recent years, LLM Multi-Agent systems have garnered widespread attention for their collaborative approach to solving complex problems. However, it’s a common scenario for these systems to fail at a task despite a flurry of activity. This leaves developers with a critical question: which agent, at what point, was responsible for the failure? Sifting through vast interaction logs to pinpoint the root cause feels like finding a needle in a haystack—a time-consuming and labor-intensive effort.
 
This is a familiar frustration for developers. In increasingly complex Multi-Agent systems, failures are not only common but also incredibly difficult to diagnose due to the autonomous nature of agent collaboration and long information chains. Without a way to quickly identify the source of a failure, system iteration and optimization grind to a halt.
 
To address this challenge, researchers from Penn State University and Duke University, in collaboration with institutions including Google DeepMind, have introduced the novel research problem of “Automated Failure Attribution.” They have constructed the first benchmark dataset for this task, Who&When, and have developed and evaluated several automated attribution methods. This work not only highlights the complexity of the task but also paves a new path toward enhancing the reliability of LLM Multi-Agent systems.

The paper has been accepted as a Spotlight presentation at the top-tier machine learning conference, ICML 2025, and the code and dataset are now fully open-source.

Paper：https://arxiv.org/pdf/2505.00212
Code：https://github.com/mingyin1/Agents_Failure_Attribution
Dataset：https://huggingface.co/datasets/Kevin355/Who_and_When
 
 
Research Background and Challenges
LLM-driven Multi-Agent systems have demonstrated immense potential across many domains. However, these systems are fragile; errors by a single agent, misunderstandings between agents, or mistakes in information transmission can lead to the failure of the entire task.

Currently, when a system fails, developers are often left with manual and inefficient methods for debugging:
Manual Log Archaeology : Developers must manually review lengthy interaction logs to find the source of the problem.
Reliance on Expertise : The debugging process is highly dependent on the developer’s deep understanding of the system and the task at hand.
 
This “needle in a haystack” approach to debugging is not only inefficient but also severely hinders rapid system iteration and the improvement of system reliability. There is an urgent need for an automated, systematic method to pinpoint the cause of failures, effectively bridging the gap between “evaluation results” and “system improvement.”

Core Contributions
This paper makes several groundbreaking contributions to address the challenges above:
1. Defining a New Problem: The paper is the first to formalize “automated failure attribution” as a specific research task. This task is defined by identifying the failure-responsible agent and the decisive error step that led to the task’s failure.
2. Constructing the First Benchmark Dataset: Who&When : This dataset includes a wide range of failure logs collected from 127 LLM Multi-Agent systems, which were either algorithmically generated or hand-crafted by experts to ensure realism and diversity. Each failure log is accompanied by fine-grained human annotations for:
Who: The agent responsible for the failure.
When: The specific interaction step where the decisive error occurred.
Why: A natural language explanation of the cause of the failure.

3. Exploring Initial “Automated Attribution” Methods : Using the Who&When dataset, the paper designs and assesses three distinct methods for automated failure attribution:
– All-at-Once: This method provides the LLM with the user query and the complete failure log, asking it to identify the responsible agent and the decisive error step in a single pass. While cost-effective, it may struggle to pinpoint precise errors in long contexts.
– Step-by-Step: This approach mimics manual debugging by having the LLM review the interaction log sequentially, making a judgment at each step until the error is found. It is more precise at locating the error step but incurs higher costs and risks accumulating errors.
– Binary Search: A compromise between the first two methods, this strategy repeatedly divides the log in half, using the LLM to determine which segment contains the error. It then recursively searches the identified segment, offering a balance of cost and performance.

Experimental Results and Key Findings 
Experiments were conducted in two settings: one where the LLM knows the ground truth answer to the problem the Multi-Agent system is trying to solve (With Ground Truth) and one where it does not (Without Ground Truth). The primary model used was GPT-4o, though other models were also tested. The systematic evaluation of these methods on the Who&When dataset yielded several important insights:
– A Long Way to Go: Current methods are far from perfect. Even the best-performing single method achieved an accuracy of only about 53.5% in identifying the responsible agent and a mere 14.2% in pinpointing the exact error step. Some methods performed even worse than random guessing, underscoring the difficulty of the task.
– No “All-in-One” Solution: Different methods excel at different aspects of the problem. The All-at-Once method is better at identifying “Who,” while the Step-by-Step method is more effective at determining “When.” The Binary Search method provides a middle-ground performance.
 

– Hybrid Approaches Show Promise but at a High Cost: The researchers found that combining different methods, such as using the All-at-Once approach to identify a potential agent and then applying the Step-by-Step method to find the error, can improve overall performance. However, this comes with a significant increase in computational cost.

– State-of-the-Art Models Struggle: Surprisingly, even the most advanced reasoning models, like OpenAI o1 and DeepSeek R1, find this task challenging.- This h…

Konten dipersingkat otomatis.

🔗 Sumber: syncedreview.com

📌 MAROKO133 Update ai: The AI that scored 95% — until consultants learned it was A

Presented by SAP

When SAP ran a quiet internal experiment to gauge consultant attitudes toward AI, the results were striking. Five teams were asked to validate answers to more than 1,000 business requirements completed by SAP’s AI co-pilot, Joule for Consultants — a workload that would normally take several weeks.

Four teams were told the analysis had been completed by junior interns fresh out of school. They reviewed the material, found it impressive, and rated the work about 95% accurate.

The fifth team was told the very same answers had come from AI.

They rejected almost everything.

Only when asked to validate each answer one by one did they discover that the AI was, in fact, highly accurate — surfacing detailed insights the consultants had initially dismissed. The overall accuracy? Again, about 95%.

“The lesson learned here is that we need to be very cautious as we introduce AI — especially in how we communicate with senior consultants about its possibilities and how to integrate it into their workflows,” says Guillermo B. Vazquez Mendez, chief architect, RI business transformation and architecture, SAP America Inc.

The experiment has since become a revealing starting point for SAP’s push toward the consultant of 2030: a practitioner who is deeply human, enabled by AI, and no longer weighed down by the technical grunt work of the past.

Overcoming AI skepticism

Resistance isn’t surprising, Vazquez notes. Consultants with two or three decades of experience carry enormous institutional knowledge — and an understandable degree of caution.

But AI copilots like Joule for Consultants are not replacing expertise. They’re amplifying it.

“What Joule really does is make their very expensive time far more effective,” Vazquez says. “It removes the clerical work, so they can focus on turning out high-quality answers in a fraction of the time.”

He emphasizes this message constantly: “AI is not replacing you. It’s a tool for you. Human oversight is always required. But now, instead of spending your time looking for documentation, you’re gaining significant time and boosting the effectiveness and detail of your answers.”

The consultant time-shift: from tech execution to business insight

Historically, consultants spent about 80% of their time understanding technical systems — how processes run, how data flows, how functions execute. Customers, by contrast, spend 80% of their time focused on their business.

That mismatch is exactly where Joule steps in.

“There’s a gap there — and the bridge is AI,” Vazquez says. “It flips the time equation, enabling consultants to invest more of their energy in understanding the customer’s industry and business goals. AI takes on the heavy technical lift, so consultants can focus on driving the right business outcomes.”

Bringing new consultants up to speed

AI is also transforming how new hires learn.

“We’re excited to see Joule acting as a bridge between senior consultants, who are adapting more slowly, and interns and new consultants who are already technically savvy,” Vazquez says.

Junior consultants ramp up faster because Joule helps them operate independently. Seniors, meanwhile, engage where their insight matters most.

This is also where many consultants learn the fundamentals of today’s AI copilots. Much of the work depends on prompt engineering — for instance, instructing Joule to act as a senior chief technology architect specializing in finance and SAP S/4HANA 2023, then asking it to analyze business requirements and deliver the output as tables or PowerPoint slides.

Once they grasp how to frame prompts, consultants consistently get higher-quality, more structured answers.

New architects are also able to communicate more clearly with their more experienced counterparts. They know what they don’t know and can ask targeted questions, which makes mentorship far smoother. It’s created a real synergy, Vazquez adds — senior consultants see how quickly new hires are adapting and learning with AI, and that momentum encourages them to keep pace and adopt the technology themselves.

Looking ahead to the future of AI copilots

“We’re still in the baby steps of AI — we’re toddlers,” Vazquez says. “Right now, copilots depend on prompt engineering to get good answers. The better you prompt, the better the answer you get.”

But that represents only the earliest phase of what these systems will eventually do. As copilots mature, they’ll move beyond responding to prompts and start interpreting entire business processes — understanding the sequence of steps, identifying where human intervention is needed, and spotting where an AI agent could take over. That shift is what leads directly into agentic AI.

SAP’s depth of process knowledge is what makes that evolution possible. The company has mapped more than 3,500 business processes across industries — a repository Vazquez calls “some of the most valuable, rigorously tested processes developed in the last 50 years.” Every day, SAP systems support roughly $7.3 trillion in global commerce, giving these emerging AI agents a rich foundation to navigate and reason over.

“With that level of process insight and data, we can take a real leap forward,” he says, “equipping our consultants with agentic AI that can solve complex challenges and push us toward increasingly autonomous systems.”

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact [email protected].

🔗 Sumber: venturebeat.com